File size: 3,114 Bytes
079c32c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 |
from easydict import EasyDict
spaceinvaders_impala_config = dict(
exp_name='impala_log/spaceinvaders_impala_seed0',
env=dict(
collector_env_num=8,
evaluator_env_num=8,
n_evaluator_episode=8,
stop_value=10000000000,
env_id='SpaceInvadersNoFrameskip-v4',
#'ALE/SpaceInvaders-v5' is available. But special setting is needed after gym make.
frame_stack=4,
# manager=dict(shared_memory=False, )
),
policy=dict(
cuda=True,
# (int) the trajectory length to calculate v-trace target
unroll_len=32,
random_collect_size=500,
model=dict(
obs_shape=[4, 84, 84],
action_shape=6,
encoder_hidden_size_list=[128, 128, 256, 256],
critic_head_hidden_size=256,
critic_head_layer_num=3,
actor_head_hidden_size=256,
actor_head_layer_num=3,
),
learn=dict(
# (int) collect n_sample data, train model update_per_collect times
# here we follow impala serial pipeline
update_per_collect=2, # update_per_collect show be in [1, 10]
# (int) the number of data for a train iteration
batch_size=128,
grad_clip_type='clip_norm',
clip_value=5,
learning_rate=0.0006,
# (float) loss weight of the value network, the weight of policy network is set to 1
value_weight=0.5,
# (float) loss weight of the entropy regularization, the weight of policy network is set to 1
entropy_weight=0.01,
# (float) discount factor for future reward, defaults int [0, 1]
discount_factor=0.99,
# (float) additional discounting parameter
lambda_=0.95,
# (float) clip ratio of importance weights
rho_clip_ratio=1.0,
# (float) clip ratio of importance weights
c_clip_ratio=1.0,
# (float) clip ratio of importance sampling
rho_pg_clip_ratio=1.0,
),
collect=dict(
# (int) collect n_sample data, train model n_iteration times
n_sample=16,
collector=dict(collect_print_freq=1000, ),
),
eval=dict(evaluator=dict(eval_freq=500, )),
other=dict(replay_buffer=dict(replay_buffer_size=100000, sliced=True), ),
),
)
spaceinvaders_impala_config = EasyDict(spaceinvaders_impala_config)
main_config = spaceinvaders_impala_config
spaceinvaders_impala_create_config = dict(
env=dict(
type='atari',
import_names=['dizoo.atari.envs.atari_env'],
),
env_manager=dict(type='subprocess'),
policy=dict(type='impala'),
replay_buffer=dict(type='naive'),
)
spaceinvaders_impala_create_config = EasyDict(spaceinvaders_impala_create_config)
create_config = spaceinvaders_impala_create_config
if __name__ == '__main__':
# or you can enter ding -m serial -c spaceinvaders_impala_config.py -s 0
from ding.entry import serial_pipeline
serial_pipeline((main_config, create_config), seed=0)
|